Search

Background: cystic fibrosis (CF) is a life-threatening genetic disease, affecting around 10 500 people in the UK. Precision medicines have been developed to treat specific CF-gene mutations. The newest, elexacaftor/tezacaftor/ivacaftor (ELEX/TEZ/IVA), has been found to be highly effective in randomised controlled trials (RCTs) and became available to a large proportion of UK CF patients in 2020. Understanding the potential health economic impacts of ELEX/TEZ/IVA is vital to planning service provision.Methods: we combined observational UK CF Registry data with RCT results to project the impact of ELEX/TEZ/IVA on total days of intravenous (IV) antibiotic treatment at a population level. Registry data from 2015 to 2017 were used to develop prediction models for IV days over a 1-year period using several predictors, and to estimate 1-year population total IV days based on standards of care pre-ELEX/TEZ/IVA. We considered two approaches to imposing the impact of ELEX/TEZ/IVA on projected outcomes using effect estimates from RCTs: approach 1 based on effect estimates on FEV1% and approach 2 based on effect estimates on exacerbation rate.Results: ELEX/TEZ/IVA is expected to result in significant reductions in population-level requirements for IV antibiotics of 16.1% (~17 800 days) using approach 1 and 43.6% (~39 500 days) using approach 2. The two approaches require different assumptions. Increased understanding of the mechanisms through which ELEX/TEZ/IVA acts on these outcomes would enable further refinements to our projections.Conclusions: this work contributes to increased understanding of the changing healthcare needs of people with CF and illustrates how Registry data can be used in combination with RCT evidence to estimate population-level treatment impacts.

Inverse probability of treatment weighting can account for confounding under a number of assumptions, including that of no unmeasured confounding. A recent simulation study proposed a bootstrap bias correction, apparently demonstrating good performance in removing bias due to unmeasured confounding. We revisited the simulations, finding no evidence of bias reduction. Journal of Statistical Research 2021, Vol. 55, No. 2, pp. 293-297

BackgroundThe inappropriate use of antibiotics is understood to contribute to antimicrobial resistance. Oral antibiotics are regularly used to treat moderate-to-severe acne vulgaris. In practice, we do not know the typical length of oral antibiotic treatment courses for acne in routine primary care and what proportion of people receive more than one course of treatment following a new acne diagnosis.ObjectivesTo describe how oral antibiotics are prescribed for acne over time in UK primary care.MethodsWe conducted a descriptive longitudinal drug utilization study using routinely collected primary care data from the Clinical Practice Research Datalink GOLD (2004–2019). We included individuals (8–50 years) with a new acne diagnosis recorded between 1 January 2004 and 31 July 2019.ResultsWe identified 217 410 people with a new acne diagnosis. The median age was 17 years [interquartile range (IQR) 15–25] and median follow-up was 4.3 years (IQR 1.9–7.6). Among people with a new acne diagnosis, 96 703 (44.5%) received 248 560 prescriptions for long-term oral antibiotics during a median follow-up of 5.3 years (IQR 2.8–8.5). The median number of continuous courses of antibiotic therapy (≥ 28 days) per person was four (IQR 2–6). The majority (n = 59 010, 61.0%) of first oral antibiotic prescriptions in those with a recorded acne diagnosis were between the ages of 12 and 18. Most (n = 71 544, 74.0%) first courses for oral antibiotics were for between 28 and 90 days. The median duration of the first course of treatment was 56 days (IQR 50–93 days) and 18 127 (18.7%) of prescriptions of ≥ 28 days were for < 6 weeks. Among people who received a first course of oral antibiotic for ≥ 28 days, 56 261 (58.2%) received a second course after a treatment gap of ≥ 28 days. The median time between first and second courses was 135 days (IQR 67–302). The cumulative duration of exposure to oral antibiotics during follow-up was 255 days (8.5 months).ConclusionsFurther work is needed to understand the consequences of using antibiotics for shorter periods than recommended. Suboptimal treatment duration may result in reduced clinical effectiveness or repeated exposures, potentially contributing to antimicrobial resistance.

Background Subsequent epidemic waves have already emerged in many countries and in the absence of highly effective preventive and curative options, the role of patient characteristics on the development of outcomes needs to be thoroughly examined, especially in middle-east countries where such epidemiological studies are lacking. There is a huge pressure on the hospital services and in particular, on the Intensive Care Units (ICU). Describing the need for critical care as well as the chance of being discharged from hospital according to patient characteristics, is essential for a more efficient hospital management. The objective of this study is to describe the probabilities of admission to the ICU and the probabilities of hospital discharge among positive COVID-19 patients according to demographics and comorbidities recorded at hospital admission. Methods A prospective cohort study of all patients with COVID-19 found in the Electronic Medical Records of Jaber Al-Ahmad Al-Sabah Hospital in Kuwait was conducted. The study included 3995 individuals (symptomatic and asymptomatic) of all ages who tested positive from February 24th to May 27th, 2020, out of which 315 were treated in the ICU and 3619 were discharged including those who were transferred to a different healthcare unit without having previously entered the ICU. A competing risk analysis considering two events, namely, ICU admission and hospital discharge using flexible hazard models was performed to describe the association between event-specific probabilities and patient characteristics. Results Results showed that being male, increasing age and comorbidities such as chronic kidney disease (CKD), asthma or chronic obstructive pulmonary disease and weakened immune system increased the risk of ICU admission within 10 days of entering the hospital. CKD and weakened immune system decreased the probabilities of discharge in both females and males however, the age-related pattern differed by gender. Diabetes, which was the most prevalent comorbid condition, had only a moderate impact on both probabilities (18% overall) in contrast to CKD which had the largest effect, but presented only in 7% of those admitted to ICU and in 1% of those who got discharged. For instance, within 5 days a 50-year-old male had 19% (95% C.I.: [15,23]) probability of entering the ICU if he had none of these comorbidities, yet this risk jumped to 31% (95% C.I.: [20,46]) if he had also CKD, and to 27% in the presence of asthma/COPD (95% C.I.: [19,36]) or of weakened immune system (95% C.I.: [16,42]). Conclusions This study provides useful insight in describing the probabilities of ICU admission and hospital discharge according to age, gender, and comorbidities among confirmed COVID-19 cases in Kuwait. A web-tool is also provided to allow the user to estimate these probabilities for any combination of these covariates. These probabilities enable deeper understanding of the hospital demand according to patient characteristics which is essential to hospital management and useful for developing a vaccination strategy.

Background Avoiding “causal” language with observational study designs is common publication practice, often justified as being a more cautious approach to interpretation. Objectives We aimed to i) estimate the degree to which causality was implied by both the language linking exposures to outcomes and by action recommendations in the high-profile health literature, ii) examine disconnects between language and recommendations, iii) identify which linking phrases were most common, and iv) generate estimates by which these phrases imply causality. Methods We identified 18 of the most prominent general medical/public health/epidemiology journals, and searched and screened for articles published from 2010 to 2019 that investigated exposure/outcome pairs until we reached 65 non-RCT articles per journal (n=1,170). Two independent reviewers and an arbitrating reviewer rated the degree to which they believed causality had been implied by the language in abstracts based on written guidance. Reviewers then rated causal implications of linking words in isolation. For comparison, additional review was performed for full texts and for a secondary sample of RCTs. Results Reviewers rated the causal implication of the sentence and phrase linking the exposure and outcome as None (i.e., makes no causal implication) in 13.8%, Weak in 34.2%, Moderate in 33.2%, and Strong in 18.7% of abstracts. Reviewers identified an action recommendation in 34.2% of abstracts. Of these action recommendations, reviewers rated the causal implications as None in 5.3%, Weak in 19.0%, Moderate in 42.8% and Strong in 33.0% of cases. The implied causality of action recommendations was often higher than the implied causality of linking sentences (44.5%) or commensurate (40.3%), with 15.3% being weaker. The most common linking word root identified in abstracts was “associate” (n=535/1,170; 45.7%) (e.g. “association,” “associated,” etc). There were only 16 (1.4%) abstracts using “cause” in the linking or modifying phrases. Reviewer ratings for causal implications of word roots were highly heterogeneous, including those commonly considered non-causal. Discussion We found substantial disconnects between causal implications used to link an exposure to an outcome and the action implications made. This undercuts common assumptions about what words are often considered non-causal and that policing them eliminates causal implications. We recommend that instead of policing words, editors, researchers, and communicators should increase efforts at making research questions, as well as the potential of studies to answer them, more transparent. Summary box

Adjustment for baseline covariates in randomized trials has been shown to lead to gains in power and can protect against chance imbalances in covariates. For continuous covariates, there is a risk that the the form of the relationship between the covariate and outcome is misspecified when taking an adjusted approach. Using a simulation study focusing on individually randomized trials with small sample sizes, we explore whether a range of adjustment methods are robust to misspecification, either in the covariate–outcome relationship or through an omitted covariate–treatment interaction. Specifically, we aim to identify potential settings where G-computation, inverse probability of treatment weighting (IPTW), augmented inverse probability of treatment weighting (AIPTW) and targeted maximum likelihood estimation (TMLE) offer improvement over the commonly used analysis of covariance (ANCOVA). Our simulations show that all adjustment methods are generally robust to model misspecification if adjusting for a few covariates, sample size is 100 or larger, and there are no covariate–treatment interactions. When there is a non-linear interaction of treatment with a skewed covariate and sample size is small, all adjustment methods can suffer from bias; however, methods that allow for interactions (such as G-computation with interaction and IPTW) show improved results compared to ANCOVA. When there are a high number of covariates to adjust for, ANCOVA retains good properties while other methods suffer from under- or over-coverage. An outstanding issue for G-computation, IPTW and AIPTW in small samples is that standard errors are underestimated; they should be used with caution without the availability of small-sample corrections, development of which is needed. These findings are relevant for covariate adjustment in interim analyses of larger trials.

Balancing the distributions of the confounders across the exposure levels in an observational study through matching or weighting is an accepted method to control for confounding due to these variables when estimating the association between an exposure and outcome and to reduce the degree of dependence on certain modeling assumptions. Despite the increasing popularity in practice, these procedures cannot be immediately applied to datasets with missing values. Multiple imputation of the missing data is a popular approach to account for missing values while preserving the number of units in the dataset and accounting for the uncertainty in the missing values. However, to the best of our knowledge, there is no comprehensive matching and weighting software that can be easily implemented with multiply imputed datasets. In this paper, we review this problem and suggest a framework to map out the matching and weighting multiply imputed datasets to 5 actions as well as the best practices to assess balance in these datasets after matching and weighting. We also illustrate these approaches using a companion package for R, MatchThem., Comment: 23 Pages, 3 Figures

ObjectivesWe aimed to assess the association between multimorbidity and deprivation on short-term mortality among patients with diffuse large B-cell (DLBCL) and follicular lymphoma (FL) in England.SettingThe association of multimorbidity and socioeconomic deprivation on survival among patients diagnosed with DLBCL and FL in England between 2005 and 2013. We linked the English population-based cancer registry with electronic health records databases and estimated adjusted mortality rate ratios by multimorbidity and deprivation status. Using flexible hazard-based regression models, we computed DLBCL and FL standardised mortality risk by deprivation and multimorbidity at 1 year.ResultsOverall, 41 422 patients aged 45–99 years were diagnosed with DLBCL or FL in England during 2005–2015. Most deprived patients with FL with multimorbidities had three times higher hazard of 1-year mortality (HR: 3.3, CI 2.48 to 4.28, pConclusionsMultimorbidity, deprivation and their combination are strong and independent predictors of an increased short-term mortality risk among patients with DLBCL and FL in England. Public health measures targeting the reduction of multimorbidity among most deprived patients with DLBCL and FL are needed to reduce the short-term mortality gap.

The main purpose of many medical studies is to estimate the effects of a treatment or exposure on an outcome. However, it is not always possible to randomize the study participants to a particular treatment, therefore observational study designs may be used. There are major challenges with observational studies; one of which is confounding. Controlling for confounding is commonly performed by direct adjustment of measured confounders; although, sometimes this approach is suboptimal due to modeling assumptions and misspecification. Recent advances in the field of causal inference have dealt with confounding by building on classical standardization methods. However, these recent advances have progressed quickly with a relative paucity of computational-oriented applied tutorials contributing to some confusion in the use of these methods among applied researchers. In this tutorial, we show the computational implementation of different causal inference estimators from a historical perspective where new estimators were developed to overcome the limitations of the previous estimators (ie, nonparametric and parametric g-formula, inverse probability weighting, double-robust, and data-adaptive estimators). We illustrate the implementation of different methods using an empirical example from the Connors study based on intensive care medicine, and most importantly, we provide reproducible and commented code in Stata, R, and Python for researchers to adapt in their own observational study. The code can be accessed at https://github.com/migariane/Tutorial_Computational_Causal_Inference_Estimators.

Missing data is a common issue in research using observational studies to investigate the effect of treatments on health outcomes. When missingness occurs only in the covariates, a simple approach is to use missing indicators to handle the partially observed covariates. The missing indicator approach has been criticized for giving biased results in outcome regression. However, recent papers have suggested that the missing indicator approach can provide unbiased results in propensity score analysis under certain assumptions. We consider assumptions under which the missing indicator approach can provide valid inferences, namely, (1) no unmeasured confounding within missingness patterns; either (2a) covariate values of patients with missing data were conditionally independent of treatment or (2b) these values were conditionally independent of outcome; and (3) the outcome model is correctly specified: specifically, the true outcome model does not include interactions between missing indicators and fully observed covariates. We prove that, under the assumptions above, the missing indicator approach with outcome regression can provide unbiased estimates of the average treatment effect. We use a simulation study to investigate the extent of bias in estimates of the treatment effect when the assumptions are violated and we illustrate our findings using data from electronic health records. In conclusion, the missing indicator approach can provide valid inferences for outcome regression, but the plausibility of its assumptions must first be considered carefully.

Background Malaria is transmitted through the bite of Plasmodium‐infected adult female Anopheles mosquitoes. Ivermectin, an anti‐parasitic drug, acts by killing mosquitoes that are exposed to the drug while feeding on the blood of people (known as blood feeds) who have ingested the drug. This effect on mosquitoes has been demonstrated by individual randomized trials. This effect has generated interest in using ivermectin as a tool for malaria control. Objectives To assess the effect of community administration of ivermectin on malaria transmission. Search methods We searched the Cochrane Infectious Diseases Group (CIDG) Specialized Register, CENTRAL, MEDLINE, Embase, LILACS, Science Citation index ‐ expanded, the World Health Organization (WHO) International Clinical Trials Registry Platform, ClinicalTrials.gov, and the National Institutes of Health (NIH) RePORTER database to 14 January 2021. We checked the reference lists of included studies for other potentially relevant studies, and contacted researchers working in the field for unpublished and ongoing trials. Selection criteria We included cluster‐randomized controlled trials (cRCTs) that compared ivermectin, as single or multiple doses, with a control treatment or placebo given to populations living in malaria‐endemic areas, in the context of mass drug administration. Primary outcomes were prevalence of malaria parasite infection and incidence of clinical malaria in the community. Data collection and analysis Two review authors independently extracted data on the number of events and the number of participants in each trial arm at the time of assessment. For rate data, we noted the total time at risk in each trial arm. To assess risk of bias, we used Cochrane's RoB 2 tool for cRCTs. We documented the method of data analysis, any adjustments for clustering or other covariates, and recorded the estimate of the intra‐cluster correlation (ICC) coefficient. We re‐analysed the trial data provided by the trial authors to adjust for cluster effects. We used a Poisson mixed‐effect model with small sample size correction, and a cluster‐level analysis using the linear weighted model to adequately adjust for clustering. Main results We included one cRCT and identified six ongoing trials. The included cRCT examined the incidence of malaria in eight villages in Burkina Faso, randomized to two arms. Both trial arms received a single dose of ivermectin 150 µg/kg to 200 µg/kg, together with a dose of albendazole. The villages in the intervention arm received an additional five doses of ivermectin, once every three weeks. Children were enrolled into an active cohort, in which they were repeatedly screened for malaria infection. The primary outcome was the cumulative incidence of uncomplicated malaria in a cohort of children aged five years and younger, over the 18‐week study. We judged the study to be at high risk of bias, as the analysis did not account for clustering or correlation between participants in the same village. The study did not demonstrate an effect of Ivermectin on the cumulative incidence of uncomplicated malaria in the cohort of children over the 18‐week study (risk ratio 0.86, 95% confidence interval (CI) 0.62 to 1.17; P = 0.2607; very low‐certainty evidence). Authors' conclusions We are uncertain whether community administration of ivermectin has an effect on malaria transmission, based on one trial published to date., Plain language summary Malaria control using ivermectin What is the aim of this review? The aim of this Cochrane Review was to find out if giving the drug ivermectin to entire communities could reduce malaria transmission. We examined all relevant studies to answer this question, and found one relevant study. Key messages It is not possible to say at this point if treating an entire community with ivermectin reduces malaria. Several research studies are in progress; we anticipate they will provide more answers in the future. What was studied in the review? Malaria is a disease transmitted to humans through the bite of mosquitoes infected with Plasmodium parasites. It results in nearly half a million deaths every year. Ivermectin is a drug that is given to whole communities to control the parasites that are responsible for elephantiasis and river blindness. It has been observed that ivermectin can kill mosquitoes when they feed on the blood of people who have taken this medication. Therefore, it is believed that by giving this drug to whole communities, it will kill many mosquitoes, and could reduce malaria transmission. In this review, we assessed whether treating entire communities with ivermectin would reduce malaria transmission. We looked for studies from different sources, and only included studies that took place in communities with malaria, and that randomly assigned groups of people to ivermectin or a control, which could be a placebo or standard community drug treatments. We wanted to know if the treatment influenced the occurrence of malaria in the community. What are the main results of the review? One study met the inclusion criteria. This study included eight villages in Burkina Faso, which were randomly assigned to receive ivermectin or a control. All villages received ivermectin, as part of the scheduled control of lymphatic filariasis. In addition, the treatment villages received five more doses of ivermectin, once every three weeks. The effect of ivermectin on malaria was measured in children younger than five years of age. In these children, the treatment did not show a notable difference in the presence of malaria between the treatment and control groups (very low‐certainty evidence). Therefore, it is not possible to say at this point if the treatment of entire communities with ivermectin has an effect on reducing malaria. Several studies are currently ongoing; we anticipate they will provide more answers in the future. How up‐to‐date is this review? We searched for studies published up to 14 January 2021.

It was highlighted that in the original article [1] the Additional file 1 was missing. This Correction article includes the Additional file 1. The original article has been updated.

AIMS: To determine whether initiation of treatment with angiotensin converting enzyme inhibitors or angiotensin II receptor blockers (ACEI/ARBs) is associated with a subsequent reduction in haemoglobin in the general population. METHODS: We undertook a national cohort study over a thirteen-year period (2004 to 2016), using routine primary healthcare data from the United Kingdom Clinical Practice Research Datalink. We compared ACEI/ARB initiation with calcium channel blocker (CCB) initiation, to minimise confounding by indication. We included all first ACEI/ARB or CCB prescriptions in adults with at least one haemoglobin result in the 12 months before and six months after drug initiation. Our primary outcome was a ≥1g/dL haemoglobin reduction in the six months after drug initiation. RESULTS: We examined 146,610 drug initiation events in 136,655 patients. Haemoglobin fell by ≥1g/dL after drug initiation in 19.5% (16,936/86,652) of ACEI/ARB initiators and 15.9% (9,521/59,958) of CCB initiators. The adjusted odds ratio of a ≥1g/dL haemoglobin reduction in ACEI/ARB initiators vs. CCB initiators was 1.15 (95%CI 1.12-1.19). CONCLUSIONS: ACEI/ARBs are associated with a modest increase in the risk of a haemoglobin reduction. For every 100 patients in our study that initiated a CCB, 16 experienced a ≥1g/dL haemoglobin decline. If the effect is causal, three additional patients would have experienced this outcome if they had received an ACEI/ARB. This may have implications for drug choice and monitoring for many patients in primary care. Further research could identify patients at higher risk of this outcome, who may benefit from closer monitoring.

In the last decade Open Science principles have been successfully advocated for and are being slowly adopted in different research communities. In response to the COVID-19 pandemic many publishers and researchers have sped up their adoption of Open Science practices, sometimes embracing them fully and sometimes partially or in a sub-optimal manner. In this article, we express concerns about the violation of some of the Open Science principles and its potential impact on the quality of research output. We provide evidence of the misuses of these principles at different stages of the scientific process. We call for a wider adoption of Open Science practices in the hope that this work will encourage a broader endorsement of Open Science principles and serve as a reminder that science should always be a rigorous process, reliable and transparent, especially in the context of a pandemic where research findings are being translated into practice even more rapidly. We provide all data and scripts at https://osf.io/renxy/.

Background The safety of restarting angiotensin-converting enzyme inhibitors (ACEI) or angiotensin II receptor blockers (ARB) after acute kidney injury (AKI) is unclear. There is concern that previous users do not restart ACEI/ARB despite ongoing indications. We sought to determine the risk of adverse events after an episode of AKI, comparing prior ACEI/ARB users who stop treatment to those who continue. Methods We conducted two parallel cohort studies in English and Swedish primary and secondary care, 2006–2016. We used multivariable Cox regression to estimate hazard ratios (HR) for hospital admission with heart failure (primary analysis), AKI, stroke, or death within 2 years after hospital discharge following a first AKI episode. We compared risks of admission between people who stopped ACEI/ARB treatment to those who were prescribed ACEI/ARB within 30 days of AKI discharge. We undertook sensitivity analyses, including propensity score-matched samples, to explore the robustness of our results. Results In England, we included 7303 people with AKI hospitalisation following recent ACEI/ARB therapy for the primary analysis. Four thousand three (55%) were classified as stopping ACEI/ARB based on no prescription within 30 days of discharge. In Sweden, we included 1790 people, of whom 1235 (69%) stopped treatment. In England, no differences were seen in subsequent risk of heart failure (HR 1.10; 95% confidence intervals (CI) 0.93–1.30), AKI (HR 0.90; 95% CI 0.77–1.05), or stroke (HR 0.99; 95% CI 0.71–1.38), but there was an increased risk of death (HR 1.27; 95% CI 1.15–1.41) in those who stopped ACEI/ARB compared to those who continued. Results were similar in Sweden: no differences were seen in risk of heart failure (HR 0.91; 95% CI 0.73–1.13) or AKI (HR 0.81; 95% CI 0.54–1.21). However, no increased risk of death was seen (HR 0.94; 95% CI 0.78–1.13) and stroke was less common in people who stopped ACEI/ARB (HR 0.56; 95% CI 0.34–0.93). Results were similar across all sensitivity analyses. Conclusions Previous ACEI/ARB users who continued treatment after an episode of AKI did not have an increased risk of heart failure or subsequent AKI compared to those who stopped the drugs.

IntroductionAntimicrobial resistance (AMR) is a global health emergency. Acne vulgaris is a highly prevalent condition and the dominant role antibiotics play in its treatment is a major concern. Antibiotics are widely used in the treatment of acne predominantly for their anti-inflammatory effect, hence their use in acne may not be optimal. Tetracyclines and macrolides are the two most common oral antibiotic classes prescribed, and their average use can extend from a few months to several years of intermittent or continuous use. The overall aim of this systematic review is to elucidate what is known about oral antibiotics for acne contributing to antibiotic treatment failure and AMR.Methods and analysisA systematic review will be conducted to address the question: What is the existing evidence that long-term oral antibiotics used to treat acne in those over 8 years of age contribute towards antibiotic treatment failure or other outcomes suggestive of the impact of AMR? We will search the following databases: Embase, MEDLINE, the Cochrane Library and Web of Science. Search terms will be developed in collaboration with a librarian by identifying keywords from relevant articles and by undertaking pilot searches. Randomised controlled trials, cohort and case-controlled studies conducted in any healthcare setting and published in any language will be included. The searches will be re-run prior to final analyses to capture the recent literature. The Cochrane tool for bias assessment in randomised trials and ROBINS-I for the assessment of bias in non-randomised studies will be used to assess the risk of bias of included studies. GRADE will be used to make an overall assessment of the quality of evidence. A meta-analysis will be undertaken of the outcome measures if the individual studies are sufficiently homogeneous. If a meta-analysis is not possible, a qualitative assessment will be presented as a narrative review.Ethics and disseminationEthical approval is not required for this systematic-review. The results will be published in a peer-reviewed journal and any deviations from the protocol will be clearly documented in the published manuscript of the full systematic-review.PROSPERO registration numberCRD42019121738.

In time-to-event settings, g-computation and doubly robust estimators are based on discrete-time data. However, many biological processes are evolving continuously over time. In this paper, we extend the g-computation and the doubly robust standardisation procedures to a continuous-time context. We compare their performance to the well-known inverse-probability-weighting (IPW) estimator for the estimation of the hazard ratio and restricted mean survival times difference, using a simulation study. Under a correct model specification, all methods are unbiased, but g-computation and the doubly robust standardisation are more efficient than inverse probability weighting. We also analyse two real-world datasets to illustrate the practical implementation of these approaches. We have updated the R package RISCA to facilitate the use of these methods and their dissemination., Accepted for publication in Statistical Methods in Medical Research, 16 pages, including 4 figures and 2 tables

Acquiring real-world evidence is crucial to support health policy, but observational studies are prone to serious biases. An approach was recently proposed to overcome confounding and immortal-time biases within the emulated trial framework. This tutorial provides a step-by-step description of the design and analysis of emulated trials, as well as R and Stata code, to facilitate its use in practice. The steps consist in: (i) specifying the target trial and inclusion criteria; (ii) cloning patients; (iii) defining censoring and survival times; (iv) estimating the weights to account for informative censoring introduced by design; and (v) analysing these data. These steps are illustrated with observational data to assess the benefit of surgery among 70–89-year-old patients diagnosed with early-stage lung cancer. Because of the severe unbalance of the patient characteristics between treatment arms (surgery yes/no), a naïve Kaplan-Meier survival analysis of the initial cohort severely overestimated the benefit of surgery on 1-year survival (22% difference), as did a survival analysis of the cloned dataset when informative censoring was ignored (17% difference). By contrast, the estimated weights adequately removed the covariate imbalance. The weighted analysis still showed evidence of a benefit, though smaller (11% difference), of surgery among older lung cancer patients on 1-year survival. Complementing the CERBOT tool, this tutorial explains how to proceed to conduct emulated trials using observational data in the presence of immortal-time bias. The strength of this approach is its transparency and its principles that are easily understandable by non-specialists.

Marginal structural models (MSMs) are commonly used to estimate causal intervention effects in longitudinal nonrandomized studies. A common challenge when using MSMs to analyze observational studies is incomplete confounder data, where a poorly informed analysis method will lead to biased estimates of intervention effects. Despite a number of approaches described in the literature for handling missing data in MSMs, there is little guidance on what works in practice and why. We reviewed existing missing-data methods for MSMs and discussed the plausibility of their underlying assumptions. We also performed realistic simulations to quantify the bias of 5 methods used in practice: complete-case analysis, last observation carried forward, the missingness pattern approach, multiple imputation, and inverse-probability-of-missingness weighting. We considered 3 mechanisms for nonmonotone missing data encountered in research based on electronic health record data. Further illustration of the strengths and limitations of these analysis methods is provided through an application using a cohort of persons with sleep apnea: the research database of the French Observatoire Sommeil de la Fédération de Pneumologie. We recommend careful consideration of 1) the reasons for missingness, 2) whether missingness modifies the existing relationships among observed data, and 3) the scientific context and data source, to inform the choice of the appropriate method(s) for handling partially observed confounders in MSMs.

Controlling for confounding bias is crucial in causal inference. Distinct methods are currently employed to mitigate the effects of confounding bias. Each requires the introduction of a set of covariates, which remains difficult to choose, especially regarding the different methods. We conduct a simulation study to compare the relative performance results obtained by using four different sets of covariates (those causing the outcome, those causing the treatment allocation, those causing both the outcome and the treatment allocation, and all the covariates) and four methods: g-computation, inverse probability of treatment weighting, full matching and targeted maximum likelihood estimator. Our simulations are in the context of a binary treatment, a binary outcome and baseline confounders. The simulations suggest that considering all the covariates causing the outcome led to the lowest bias and variance, particularly for g-computation. The consideration of all the covariates did not decrease the bias but significantly reduced the power. We apply these methods to two real-world examples that have clinical relevance, thereby illustrating the real-world importance of using these methods. We propose an R package RISCA to encourage the use of g-computation in causal inference.

IntroductionIdentifying design features and implementation strategies to optimise community health worker (CHW) programmes is important in the context of mixed results at scale. We systematically reviewed evidence of the effects of proactive case detection by CHWs in low-income and middle-income countries (LMICs) on mortality, morbidity and access to care for common childhood illnesses.MethodsPublished studies were identified via electronic databases from 1978 to 2017. We included randomised and non-randomised controlled trials, controlled before–after studies and interrupted time series studies, and assessed their quality for risk of bias. We reported measures of effect as study investigators reported them, and synthesised by outcomes of mortality, disease prevalence, hospitalisation and access to treatment. We calculated risk ratios (RRs) as a principal summary measure, with CIs adjusted for cluster design effect.ResultsWe identified 14 studies of 11 interventions from nine LMICs that met inclusion criteria. They showed considerable diversity in intervention design and implementation, comparison, outcomes and study quality, which precluded meta-analysis. Proactive case detection may reduce infant mortality (RR: 0.52–0.94) and increase access to effective treatment (RR: 1.59–4.64) compared with conventional community-based healthcare delivery (low certainty evidence). It is uncertain whether proactive case detection reduces mortality among children under 5 years (RR: 0.04–0.80), prevalence of infectious diseases (RR: 0.06–1.02), hospitalisation (RR: 0.38–1.26) or increases access to prompt treatment (RR: 1.00–2.39) because the certainty of this evidence is very low.ConclusionProactive case detection may provide promising benefits for child health, but evidence is insufficient to draw conclusions. More research is needed on proactive case detection with rigorous study designs that use standardised outcomes and measurement methods, and report more detail on complex intervention design and implementation.PROSPERO registration numberCRD42017074621.

ObjectivesTo assess the effectiveness of an internet-accessed STI (e-STI) testing and results service on testing uptake among young adults (16–30 years) who have never tested for STIs in London, England.MethodsWe conducted secondary analyses on data from a randomised controlled trial. In the trial, participants were randomly allocated to receive a text message with the web link of an e-STI testing and results service (intervention group) or a text message with the link of a website listing the locations, contact details and websites of seven local sexual health clinics (control group). We analysed a subsample of 528 trial participants who reported never testing for STIs at baseline. Outcomes were self-reported STI testing at 6 weeks, verified by patient record checks, and time from randomisation to completion of an STI test.ResultsUptake of STI testing among ‘never testers’ almost doubled. At 6 weeks, 45.3% of the intervention completed at least one test (chlamydia, gonorrhoea, syphilis and HIV), compared with 24.1% of the control (relative risk [RR] 1.88, 95% CI 1.47 to 2.40, pConclusions e-STI testing increased uptake of STI testing and reduced time to test among a young population of ‘never testers’ recruited in community settings. Although encouraging, questions remain on how best to manage the additional demand generated by e-STI testing in a challenging funding environment. Larger studies are required to assess the effects later in the cascade of care, including STI diagnoses and cases treated.

BackgroundAntimicrobial resistance (AMR) is a global health priority. Acne vulgaris is a common skin condition for which antibiotic use ranges from a few months to years of daily exposure.AimTo systemically search for and synthesise evidence on the risk of treatment-resistant infections, and other evidence of AMR, following long-term oral antibiotic use for acne.Design & settingIn this systematic review, a literature search was carried out using the databases Embase, MEDLINE, Cochrane, and Web of Science. They were searched using MeSH, Emtree, or other relevant terms, and followed a pre-registered protocol.MethodSearch strategies were developed with a librarian and undertaken in July 2019. All searches date from database inception. The primary outcome was antibiotic treatment failure or infection caused by a resistant organism. Secondary outcomes included detection of resistant organisms without an infection, rate of infection, or changes to flora.ResultsA total of 6996 records were identified. Seventy-three full-text articles were shortlisted for full review, of which five were included. Two investigated rates of infection, and three resistance or changes to microbial flora. Three studies had 35 or fewer participants (range 20–118 496). Three studies had a serious or high risk of bias, one moderate, and one a low risk of bias. Weak evidence was found for an association between antibiotic use for acne and subsequent increased rates of upper respiratory tract infections and pharyngitis.ConclusionThere is a lack of high quality evidence on the relationship between oral antibiotics for acne treatment and subsequent AMR sequelae. This needs to be urgently addressed with rigorously conducted studies.

Introduction Dans les etudes observationnelles, les methodes basees sur le score de propension connaissent une popularite croissante pour estimer l’effet causal populationnel d’une exposition, d’un traitement ou d’une autre sorte d’intervention. Parmi ces differentes methodes, l’« Inverse-Probability-Weighting (IPW) [1] est decrite comme etant la plus fiable. Egalement basee sur le modele causal a resultats potentiels, la « g-computation » (GC) [2] constitue une alternative prometteuse aux scores de propension. Cependant, aucune etude de simulation n’a compare ces deux methodes dans un contexte de survie en presence de censure a droite. Methodes Nous proposons une etude de simulation comparant la GC et l’IPW en presence de divers taux de censure a droite. Sachant que l’estimateur le plus courant dans ce cadre, le hazard ratio, pose question pour tirer des conclusions causales claires, nous avons choisi de nous interesser a un estimateur absolu : la difference de temps de survie moyen restreint a un horizon temporel (RMST) [3] . Puisqu’il n’y a pas de consensus dans la litterature sur quelle ensemble de covariables ajuster dans ce contexte, nous investiguons egalement deux ensembles : toutes les variables causant l’evenement a l’inclusion ou seulement le sous-vecteur causant en plus l’exposition a l’inclusion. Resultats Dans notre etude de simulation, nous rapportons que la GC et l’IPW estiment de facon non-biaisee la difference de RMST, mais la GC est associee a une plus forte puissance statistique particulierement avec un ajustement sur les variables causant l’evenement. Ces resultats ont ete illustres par deux applications sur des donnees de vie reelles tirees d’etudes recemment publiees dans le domaine de la sclerose en plaques et de la transplantation renale. Conclusion Nous illustrons ici la puissance superieure de la GC par rapport au score de propension pour estimer un effet causal marginal absolu en presence de confusion non-dependante du temps. Notre travail s’inscrit donc dans la continuite d’une litterature emergente qui remet en question l’utilisation accrue des scores de propension par les statisticiens appliques pour l’estimation des effets causaux marginaux. Nous proposons en sus un package R dans lesquels la GC a ete implementee pour encourager son utilisation en inference causale.

Inverse probability of treatment weighting is a popular propensity score-based approach to estimate marginal treatment effects in observational studies at risk of confounding bias. A major issue when estimating the propensity score is the presence of partially observed covariates. Multiple imputation is a natural approach to handle missing data on covariates: covariates are imputed and a propensity score analysis is performed in each imputed dataset to estimate the treatment effect. The treatment effect estimates from each imputed dataset are then combined to obtain an overall estimate. We call this method MIte. However, an alternative approach has been proposed, in which the propensity scores are combined across the imputed datasets (MIps). Therefore, there are remaining uncertainties about how to implement multiple imputation for propensity score analysis: (a) should we apply Rubin’s rules to the inverse probability of treatment weighting treatment effect estimates or to the propensity score estimates themselves? (b) does the outcome have to be included in the imputation model? (c) how should we estimate the variance of the inverse probability of treatment weighting estimator after multiple imputation? We studied the consistency and balancing properties of the MIte and MIps estimators and performed a simulation study to empirically assess their performance for the analysis of a binary outcome. We also compared the performance of these methods to complete case analysis and the missingness pattern approach, which uses a different propensity score model for each pattern of missingness, and a third multiple imputation approach in which the propensity score parameters are combined rather than the propensity scores themselves (MIpar). Under a missing at random mechanism, complete case and missingness pattern analyses were biased in most cases for estimating the marginal treatment effect, whereas multiple imputation approaches were approximately unbiased as long as the outcome was included in the imputation model. Only MIte was unbiased in all the studied scenarios and Rubin’s rules provided good variance estimates for MIte. The propensity score estimated in the MIte approach showed good balancing properties. In conclusion, when using multiple imputation in the inverse probability of treatment weighting context, MIte with the outcome included in the imputation model is the preferred approach.

Background Cluster randomized trials (CRTs) and individually randomized trials (IRTs) are often pooled together in meta-analyses (MAs) of randomized trials. However, the potential systematic differences in intervention effect estimates between these two trial types has never been investigated. Therefore, we conducted a meta-epidemiological study comparing intervention effect estimates between CRTs and IRTs. Methods All Cochrane MAs including at least one CRT and one IRT, published between 1 January 2010 and 31 December 2014, were included. For each MA, we estimated a ratio of odds ratios (ROR) for binary outcomes or a difference of standardized differences (DSMD) for continuous outcomes, where less than 1 (or 0, respectively) indicated a greater intervention effect estimate with CRTs. Results Among 1301 screened reviews, we selected 121 MAs, of which 76 had a binary outcome and 45 had a continuous outcome. For binary outcomes, intervention effect estimates did not differ between CRTs and IRTs [ROR 1.00, 95% confidence interval (0.93 to 1.08)]. Subgroup and adjusted analyses led to consistent results. For continuous outcomes, the DSMD was 0.13 (0.06 to 0.19). It was lower for MAs with a pharmacological intervention [-0.03, (-0.12 to 0.07)], an objective outcome [0.05, (-0.08 to 0.17)] or after adjusting for trial size [0.06, (-0.01 to 0.15)]. Conclusion For binary outcomes, CRTs and IRTs can safely be pooled in MAs because of an absence of systematic differences between effect estimates. For continuous outcomes, the results were less clear although accounting for trial sample sizes led to a non-significant difference. More research is needed for continuous outcomes and, meanwhile, MAs should be completed with subgroup analyses (CRTs vs IRTs).

Electronic health records are a valuable data source for investigating health-related questions, and propensity score analysis has become an increasingly popular approach to address confounding bias in such investigations. However, because electronic health records are typically routinely recorded as part of standard clinical care, there are often missing values, particularly for potential confounders. In our motivating study-using electronic health records to investigate the effect of renin-angiotensin system blockers on the risk of acute kidney injury-two key confounders, ethnicity and chronic kidney disease stage, have 59% and 53% missing data, respectively. The missingness pattern approach (MPA), a variant of the missing indicator approach, has been proposed as a method for handling partially observed confounders in propensity score analysis. In the MPA, propensity scores are estimated separately for each missingness pattern present in the data. Although the assumptions underlying the validity of the MPA are stated in the literature, it can be difficult in practice to assess their plausibility. In this article, we explore the MPA's underlying assumptions by using causal diagrams to assess their plausibility in a range of simple scenarios, drawing general conclusions about situations in which they are likely to be violated. We present a framework providing practical guidance for assessing whether the MPA's assumptions are plausible in a particular setting and thus deciding when the MPA is appropriate. We apply our framework to our motivating study, showing that the MPA's underlying assumptions appear reasonable, and we demonstrate the application of MPA to this study.

Background: Cluster randomized trials (CRTs) are increasingly used to assess the effectiveness of health interventions. Three main analysis approaches are: cluster-level analyses, mixed-models and generalized estimating equations (GEEs). Mixed models and GEEs can lead to inflated type I error rates with a small number of clusters, and numerous small-sample corrections have been proposed to circumvent this problem. However, the impact of these methods on power is still unclear. Methods: We performed a simulation study to assess the performance of 12 analysis approaches for CRTs with a continuous outcome and 40 or fewer clusters. These included weighted and unweighted cluster-level analyses, mixed-effects models with different degree-of-freedom corrections, and GEEs with and without a small-sample correction. We assessed these approaches across different values of the intraclass correlation coefficient (ICC), numbers of clusters and variability in cluster sizes. Results: Unweighted and variance-weighted cluster-level analysis, mixed models with degree-of-freedom corrections, and GEE with a small-sample correction all maintained the type I error rate at or below 5% across most scenarios, whereas uncorrected approaches lead to inflated type I error rates. However, these analyses had low power (below 50% in some scenarios) when fewer than 20 clusters were randomized, with none reaching the expected 80% power. Conclusions: Small-sample corrections or variance-weighted cluster-level analyses are recommended for the analysis of continuous outcomes in CRTs with a small number of clusters. The use of these corrections should be incorporated into the sample size calculation to prevent studies from being underpowered.

Introduction En Angleterre, l’âge a un fort impact sur l’acces au traitement et les perspectives de remission pour les patients atteints de cancer : ceux qui sont diagnostiques avec un cancer du poumon non-a-petites-cellules (NSCLC) apres 70 ans ont une probabilite de recevoir un traitement chirurgical reduite de 60 % par rapport aux patients de moins de 70 ans [1] . Ces patients sont generalement exclus des essais cliniques, d’ou un manque de preuves justifiant le recours a la chirurgie. L’objectif de ce travail etait de mesurer l’effet causal d’un traitement chirurgical chez les patients de plus de 70 ans, dans les six mois suivant le diagnostic de leur NSCLC, sur leur probabilite de survie a un an. Materiel et methodes L’etude a utilise les donnees des registres nationaux de cancer croisees a des donnees medicales electroniques, telles que les donnees hospitalieres (« Hospital Episode Statistics ») et les donnees des registres cliniques (« Lung Cancer Audit data »). A partir de ces donnees observationnelles, nous avons reproduit un essai clinique randomise a deux bras paralleles en incluant les patients diagnostiques en 2012 a un stade precoce, ayant un bon indice fonctionnel, sans comorbidite et âges de 70 ans ou plus au moment du diagnostic [2] . L’intervention etudiee est le traitement chirurgical dans les six mois suivant le diagnostic. Le bras temoin correspond a l’absence de chirurgie dans les six mois. Afin d’eliminer le biais d‘immortalite, les patients ont ete clones et entrent dans les deux bras au temps de reference. Dans chaque bras, les patients deviant du protocole (c’est-a-dire beneficiant de la chirurgie dans le bras temoin ou inversement, n’ayant pas beneficie de la chirurgie dans les six mois suivant le diagnostic dans le bras intervention) ont ete censures a la date de deviation. Cette censure potentiellement informative a ete prise en compte en ponderant le modele d’analyse par l’inverse de la probabilite de censure [3] . Resultats et conclusion L’utilisation de ces donnees d’observation pour reproduire un essai clinique autorise la quantification du benefice du traitement chirurgical sur la population de patients âges de 70 ans ou plus. Nous illustrerons comment cette approche permet l’estimation de l’effet causal d’un traitement chirurgical sur la probabilite de survie a un an des patients de plus de 70 ans, et mettrons en avant les enjeux statistiques de la methode d’analyse employee pour cette etude.

ObjectivesTo systematically review the quality of reporting of pilot and feasibility of cluster randomised trials (CRTs). In particular, to assess (1) the number of pilot CRTs conducted between 1 January 2011 and 31 December 2014, (2) whether objectives and methods are appropriate and (3) reporting quality.MethodsWe searched PubMed (2011–2014) for CRTs with ‘pilot’ or ‘feasibility’ in the title or abstract; that were assessing some element of feasibility and showing evidence the study was in preparation for a main effectiveness/efficacy trial. Quality assessment criteria were based on the Consolidated Standards of Reporting Trials (CONSORT) extensions for pilot trials and CRTs.ResultsEighteen pilot CRTs were identified. Forty-four per cent did not have feasibility as their primary objective, and many (50%) performed formal hypothesis testing for effectiveness/efficacy despite being underpowered. Most (83%) included ‘pilot’ or ‘feasibility’ in the title, and discussed implications for progression from the pilot to the future definitive trial (89%), but fewer reported reasons for the randomised pilot trial (39%), sample size rationale (44%) or progression criteria (17%). Most defined the cluster (100%), and number of clusters randomised (94%), but few reported how the cluster design affected sample size (17%), whether consent was sought from clusters (11%), or who enrolled clusters (17%).ConclusionsThat only 18 pilot CRTs were identified necessitates increased awareness of the importance of conducting and publishing pilot CRTs and improved reporting. Pilot CRTs should primarily be assessing feasibility, avoiding formal hypothesis testing for effectiveness/efficacy and reporting reasons for the pilot, sample size rationale and progression criteria, as well as enrolment of clusters, and how the cluster design affects design aspects. We recommend adherence to the CONSORT extensions for pilot trials and CRTs.

Background: Cluster randomised trials (CRTs) are commonly analysed using mixed-effects models or generalised estimating equations (GEEs). However, these analyses do not always perform well with the small number of clusters typical of most CRTs. They can lead to increased risk of a type I error (finding a statistically significant treatment effect when it does not exist) if appropriate corrections are not used.\ud \ud Methods: We conducted a small simulation study to evaluate the impact of using small-sample corrections for mixed-effects models or GEEs in CRTs with a small number of clusters. We then reanalysed data from TRIGGER, a CRT with six clusters, to determine the effect of using an inappropriate analysis method in practice. Finally, we reviewed 100 CRTs previously identified by a search on PubMed in order to assess whether trials were using appropriate methods of analysis. Trials were classified as at risk of an increased type I error rate if they did not report using an analysis method which accounted for clustering, or if they had fewer than 40 clusters and performed an individual-level analysis without reporting the use of an appropriate small-sample correction.\ud \ud Results: Our simulation study found that using mixed-effects models or GEEs without an appropriate correction led to inflated type I error rates, even for as many as 70 clusters. Conversely, using small-sample corrections provided correct type I error rates across all scenarios. Reanalysis of the TRIGGER trial found that inappropriate methods of analysis gave much smaller P values (P ≤ 0.01) than appropriate methods (P = 0.04–0.15). In our review, of the 99 trials that reported the number of clusters, 64 (65 %) were at risk of an increased type I error rate; 14 trials did not report using an analysis method which accounted for clustering, and 50 trials with fewer than 40 clusters performed an individual-level analysis without reporting the use of an appropriate correction.\ud \ud Conclusions: CRTs with a small or medium number of clusters are at risk of an inflated type I error rate unless appropriate analysis methods are used. Investigators should consider using small-sample corrections with mixed-effects models or GEEs to ensure valid results.\ud \ud Abbreviations: CRT, Cluster randomised trial; CI, Confidence interval; GEE, Generalised estimating equations; TRIGGER, Trial in Gastrointestinal Transfusion

In cluster randomized trials (CRTs), clusters of individuals are randomized rather than the individuals themselves. For such trials, power depends in part on the degree of similarity among responses within a cluster, which is quantified by the intaclass correlation coefficient (ICC). Thus, for a fixed sample size, power decreases with increasing ICC. In reliability studies with two observers, dichotomizing a continuous outcome variable has been shown to reduce the ICC. We checked (by a simulation study) that this property still applies to CRTs, in which cluster sizes are variable and usually greater than in reliability studies and observations (within clusters) are exchangeable. Then, in a CRT, dichotomizing a continuous outcome actually induces two antagonistic effects: decreased power because of loss of information and increased power induced by attenuation of the ICC. Therefore, we aimed to assess the impact of dichotomizing a continuous outcome on power in a CRT. We derived an analytical formula for power based on a generalized estimating equation approach after dichotomizing a continuous outcome. This theoretical result was obtained by considering equal cluster sizes, and we then assessed its accuracy (by a simulation study) in the more realistic situation of varying cluster sizes. We showed that dichotomization is associated with decreased power: attenuation of the ICC does not compensate for the loss of power induced by loss of information. Loss of power is reduced with increased initial continuous-outcome ICC and/or prevalence of success for the dichotomized outcome approaching 50%.

International audience; In cluster randomized trials, clusters of subjects are randomized rather than subjects themselves, and missing outcomes are a concern as in individual randomized trials. We assessed strategies for handling missing data when analysing cluster randomized trials with a binary outcome; strategies included complete case, adjusted complete case, and simple and multiple imputation approaches. We performed a simulation study to assess bias and coverage rate of the population-averaged intervention-effect estimate. Both multiple imputation with a random-effects logistic regression model or classical logistic regression provided unbiased estimates of the intervention effect. Both strategies also showed good coverage properties, even slightly better for multiple imputation with a random-effects logistic regression approach. Finally, this latter approach led to a slightly negatively biased intracluster correlation coefficient estimate but less than that with a classical logistic regression model strategy. We applied these strategies to a real trial randomizing households and comparing ivermectin and malathion to treat head lice.

The problem\ud A pilot or feasibility trial in preparation for a randomised trial of effectiveness is a study where part or all of a future trial is carried out on a smaller scale to see whether it can be done and whether we should proceed with it. We know reporting of pilot and feasibility studies is poor, and that these studies are particularly important when designing cluster randomised trials (CRT), in which clusters, rather than individuals, are the unit of randomisation. There are no previous reviews of the reporting quality of pilot and feasibility CRTs. The approach\ud We will systematically identify reports of pilot and feasibility CRTs published in English between 2011 and 2014. We expect around 40 studies. An electronic PubMed search, hand-search in online journals, and individual contacts will be used to identify papers. Quality assessment criteria will be based on the CONSORT extension for CRTs, and a CONSORT extension for pilot and feasibility randomised trials in the final stages of development.\ud Findings\ud This project is in progress and will be completed by September 2015. Research to date has found quality of pilot and feasibility studies is poor so we expect that here. However, there have previously been no formal guidelines against which to assess quality of pilot and feasibility studies.\ud Consequences\ud We will present recommendations for improving the conduct, analysis and reporting of these studies and expect this to improve future quality.

International audience; We examine the properties of principal scores methods to estimate the causal marginal odds ratio of an intervention for compliers in the context of a randomized controlled trial with non-compliers. The two-stage estimation approach has been proposed for a linear model by Jo and Stuart (Statistics in Medicine 2009; 28:2857-2875) under a principal ignorability (PI) assumption. Using a Monte Carlo simulation study, we compared the performance of several strategies to build and use principal score models and the robustness of the method to violations of underlying assumptions, in particular PI. Results showed that the principal score approach yielded unbiased estimates of the causal marginal log odds ratio under PI but that the method was sensitive to violations of PI, which occurs in particular when confounders are omitted from the analysis. For principal score analysis, probability weighting performed slightly better than full matching or 1:1 matching. Concerning the variables to be included in principal score models, the lowest mean squared error was generally obtained when using the true confounders. Using variables associated with the outcome only but not compliance however yielded very similar performance.

Despite randomization, selection bias may occur in cluster randomized trials. Classical multivariable regression usually allows for adjusting treatment effect estimates with unbalanced covariates. However, for binary outcomes with low incidence, such a method may fail because of separation problems. This simulation study focused on the performance of propensity score (PS)-based methods to estimate relative risks from cluster randomized trials with binary outcomes with low incidence. The results suggested that among the different approaches used (multivariable regression, direct adjustment on PS, inverse weighting on PS, and stratification on PS), only direct adjustment on the PS fully corrected the bias and moreover had the best statistical properties.

International audience; Robust evidence of the effectiveness of interventions relating to policy, practice, and organisation of healthcare often comes from well conducted cluster randomised trials. Such trials are, however, prone to recruitment bias depending on whether participants are recruited before the randomisation of clusters and whether the recruiter is blinded to the allocation status. In most cases, participants and trial staff cannot be blinded to the intervention, which might lead to performance and detection bias. Unfortunately, cluster trial reports often do not provide a clear description of the timing of trial processes and blinding, and these aspects are not covered by current reporting tools. This article proposes a graphical tool depicting the time sequence of steps and blinding status in cluster randomised trials. The tool might be helpful at both the protocol and the report writing stages to clarify the process and to help identify potential bias in cluster randomised trials.

Introduction Les methodes d’analyse par score de propension (SP) connaissent un veritable engouement pour prendre en compte les biais de confusion dans les etudes observationnelles. Un argument evoque est qu’elles permettent une interpretation causale de l’effet dans la population (effet marginal) en mimant un essai randomise controle. Dans ce travail, nous montrons que, contrairement aux idees recues, un modele de regression multivarie peut etre utilise pour obtenir une telle estimation causale avec un critere de jugement binaire. Methodes Nous etudions trois methodes : – la regression logistique multivariee ; – l’appariement sur le SP ; – la ponderation sur le SP ou inverse probability treatment weighting (IPTW). Ces methodes sont comparees pour l’estimation marginale de trois criteres : les probabilites d’evenement selon l’exposition et l’odds ratio (OR) marginal correspondant. Pour la regression multivariee, la probabilite de subir l’evenement dans la population exposee a ete estimee par la moyenne des probabilites predites individuelles de tous les patients de l’echantillon en les supposant exposes. L’OR marginal a ete calcule apres avoir estime de maniere similaire la probabilite de subir l’evenement sans etre expose. Pour l’appariement sur le SP, un appariement 1:1 a ete realise (distance – comparer l’efficacite de deux traitements (natalizumab et fingolimod) pour prevenir la rechute chez des patients atteints de sclerose en plaque de type recurrente ; – evaluer l’efficacite d’un traitement d’induction depletant pour prevenir le retard au demarrage du greffon chez des patients transplantes renaux. Resultats Les resultats des simulations montrent que la regression logistique multivariee et la ponderation IPTW sont les plus performantes en termes de biais (proches de zero) et de variance, avec une meilleure puissance pour la regression multivariee. Appliquees aux donnees reelles ces deux methodes ont des estimations marginales comparables. Une efficacite superieure du traitement natalizumab est ainsi mise en evidence pour prevenir la rechute d’une sclerose en plaque (p Conclusion La regression logistique multivariee permet une estimation de l’effet marginal d’une exposition sur la probabilite d’un evenement en presence de facteurs de confusion. Ses performances sont comparables a la ponderation sur le SP et superieures a l’appariement sur le SP. La regression multivariee presente l’avantage de pouvoir estimer a la fois l’effet marginal (populationnel) et l’effet conditionnel (patient-specifique), ce qui interesse les medecins dans leur prise de decision medicale centree sur le patient et les decideurs politiques en sante publique.